Side 30

Bygningsdesign DE FØRSTE PRODUAL PROXIMA™ -PRODUKTER www.produal.dk/pump Let the data tell. Fortsat tracting patterns, which are then interpreted. Decisionmakers can then determine how these patterns may help. To achieve optimal performance, however, it is important to start from the decision-makers and identify the most critical decisions to be taken, including the variability of their potential outcomes. The necessary knowledge to take those decisions can then be identified, and finally, the data sources that can provide it. Thus, the suggested analytical approach works best in an environment that houses: • decisions with high impact and criticality, namely early design decisions with high variability of outcome and high impact • specific performance criteria, concerning the practical implications of the decisions with regards to the targeted building performance • data from a high number of reference buildings • data in big amounts and diversity Starting from the most critical decisions usually means considering the project-specific requirements and constraints. Most of them are related and interdependent, and that dependence can be a powerful aid in data analytics. Figure 2, page 29, shows a developed dependency diagram capturing the relevant decision-making criteria, as well as those with most dependencies. Predictive models can hereby contribute further, by presenting the criticality of the decisions, the variability of outcomes and the potential impacts. As a result, the data with highest relevance to the decision-making process can be identified and prioritised in the data selection process. Figure 3. Dynamic parameters. The first source of reference data captures operational building data. Commonly collected data includes time data, energy consumption data, HVAC system operating parameters, and environmental data. These data typically track parameters that directly influence building performance and change dynamically. They can be a valuable input for simulations, HVAC system optimisation and improvement of building operation. Figure 3 shows a summary of the dynamic parameters and therefore the data types usually collected from Building Management Systems (BMS). Knowledge can also be obtained from the building design data. This data is much more static, even when taking into account versioning possibilities. Data generation at the design stage typically starts with a design brief and design model (building geometry). This data typically responds to a lot of the requirements and constraints listed earlier in the dependency diagram in figure 2, page 29. Crucial choices related to building orientation, zoning, spatial arrangement, and building materials are typical to the earliest design stages. They comprise important static information defining the character of the building, figure 4. Hidden knowledge is also available in existing simulation data and can inform the design according to the paths defined in the dependency diagram. They give an insight into the performance of the building in many potential dimensions. Yet, they are typically a lot more optimistic compared to the actual building operation. It is important to note that not all of the data is equally useful to all analytical techniques. To Figure 4. Static parameters. be useful, data needs to be carefully selected, pre-processed and transformed into appropriate analytical input. Furthermore, not each analytical technique is equally useful for any kind of dataset. Considering that the data comes from heterogeneous sources, besides categorising the data types, it is also important to select the appropriate pattern extraction and knowledge representation techniques. This allows to retrieve patterns and knowledge that is useful. Pattern recognition in operational building data When it comes to big data and large datasets, traditional analytical approaches can generate informative reports, but fail when it comes to content analysis. On the other hand, Knowledge Discovery in Databases (KDD) and Data Mining (DM) excel in analysing huge volumes of data and knowledge extraction, and can facilitate effective design space exploration. While KDD, figure 5, page 32, is the complete process of extracting useful knowledge from data, DM is the analytical part in that process that facilitates the finding of unsuspected relationships in the data. As an integral part of DM, pattern recognition is the automatic 30 HVAC 6 · 2018 The indoor climate company

Side 31

Testo 550 og 557 sæt - klar til brug Til målinger på klima- og køleanlæg > Hurtig og præcis måling af høj- og lavtryk > Samtidig beregning af overhedning og underafkøling > Inklusiv påfyldningsslanger – passer til alle kølemidler > App med visning af 60 forskellige kølemidler > Automatisk kompensation for omgivende tryk Stort udvalg af elektriske måleinstrumenter > Find dit næste elektriske måleinstrument på vores webshop. Alt indenfor multimetre, spændingstestere og amperemetre Produktnyhed 3-nyheder-i-én: Digitale kølemanifold sæt Hvilket har du brug for? HASSELLUNDEN 11A · DK-2765 SMØRUM TLF. 45 95 04 10 · WWW.BUHL-BONSOE.DK HI-FLO II XLT ePM1 60% VORES BEDSTE POSEFILTER NOGENSINDE Hi-Flo II XLT ePM1 60% (F7) Hi-Flo II XLT ePM1 60% er Camfils bedste A+ klassificerede posefilter, hvor energiforbruget er på et lavere niveau end hvad noget tidligere posefilter har kunnet præstere, samtidig med at det holder en stabil effektivitet. Sammenlignet med den forrige generations Hi-Flo posefilter sparer du cirka 166 kWh pr. filter i årligt energiforbrug. Godt for både miljø og økonomi. camfil.dk HVAC 6 · 2018 CLEAN AIR SOLUTIONS 31

    ...